Scaling the application of Artificial Intelligence in science across Antarctica

Scaling the application of Artificial Intelligence in science across Antarctica

Scaling the application of Artificial Intelligence in science across Antarctica

Lead supervisors: Dr Ben Evans

Location: AI Lab, British Antarctic Survey (BAS)

Duration: 8 weeks

Suitable undergraduate degrees: Computer Science, Mathematics

Project background


To adapt an existing machine learning approach to the problem of detection of icebergs from satellite radar imagery such that it can be deployed at scale across Antarctica.


The project provides hands-on experience of the development and scalable implementation of AI methods in polar science. An algorithm for unsupervised iceberg detection from satellite imagery has been developed and validated for the Amundsen Sea embayment (Evans et al, in prep). The approach is based on a Dirichlet Process Mixture Model, the core functionality of which does not support parallelisation. This aspect therefore represents a bottleneck for deployment at scale. Alternative, GPU enabled, implementations have been identified (e.g. using TensorFlow or Julia); the student will substitute the most appropriate within the existing Python workflow before deploying on the BAS high performance computing facility.  They will conduct tests to ensure comparable outputs with the existing method and profiling to quantify improvement in performance. They will write a report summarising the decisions made, changes implemented, and outcomes.  Once parallelised deployment is possible the student will work with GPS tracking data for a large tabular iceberg (A74) to compare satellite derived identification and drift patterns with those from the GPS, thus providing additional validation of the process in a new geographic location

The student will be based within the AI lab at BAS, Cambridge. They will be encouraged to engagement across BAS in person as far as possible through use of the social spaces and less formal meetings. The technical development activities can theoretically be carried out remotely, although again the student will be encouraged to work in BAS. They will be welcomed to the fortnightly AI lab meetings (hybrid) and made aware of other talks or meetings that may be relevant, including those surrounding the Digital Twinning efforts for Antarctica and the research vessel, to which efforts their work will make a significant contribution.

Apply now

Click here for eligibility details and how to apply